Fix a correctness bug for queries with a
LIMITon top of a
FULL OUTER JOIN.
Fix a correctness bug in how word boundaries are handled for regular expression functions when using the Joni regex engine.
Fix query failures caused by Parquet predicate pushdown for
Fix query failures when view columns were fully qualified with catalog and/or schema.
Fix query failures when creating and reading an unpartitioned Hive table within the same transaction.
Fix an issue that would disable a performance optimization if two tables with mismatching but compatible bucket counts are joined and one of the tables has a
Fix an issue that may cause procedure calls to fail if no queries were run after the server started up.
Fix an issue where the properties supported by the
ANALYZEstatement for a given connector would remain in the
system.metadata.analyze_propertiestable even after the connector was removed.
view_ownercolumn to the
JSONversion of the query plan to
Add support for creating warnings during parsing.
Add a warning for using the
current_rolereserved word as an identifier.
Add support for using the empty string as a delimiter for the
split()function. When an empty string is used as the delimiter, the string will be split into individual characters.
raptor.minimum-node-countconfiguration property. If the number of nodes in a cluster is less than the configured value, data recovery and reassignment will not run.
raptor.startup-grace-periodconfiguration property to specify the delay for the initial bucket balancer run after the coordinator startup.
Hive Connector Changes#
hive.ignore-table-bucketingconfiguration property and
ignore_table_bucketingsession property. When set to true, these properties enable reading from partitions whose bucketing scheme does not match the table bucketing scheme.
Fix an issue where the determinism check for
INSERTqueries were not run when there was a result mismatch.
Fix an issue where the checksum query was treated as the main query, and the main query was treated as part of the setup queries for
Fix query timeout enforcement by replacing local timer with the
Improve result comparison for floating point columns by using relative errors. Replace configuration property
Improve result comparison for
SELECTqueries by rewriting
CREATE TABLE ASand using checksum queries to verify the results. This eliminates the row count limit for
Reuse initial control query results for the determinism check. This reduces the maximum number of control query runs and eliminates the test query reruns.
Add support for retrying transient query failures using configuration properties
presto.backoff-scale-factor. Similar configurations prefixed with
prestoare introduced for retrying transient network failures when communicating with the coordinator. Intermediate failures are recorded and emitted in the output.
Add support for automatically resolving certain kinds of failures including exceeding the global memory limit and exceeding time limit.
Add configuration properties
checksum-timeoutto set the timeouts for metadata queries (i.e., describe queries that read table schema) and checksum queries.
source-query.table-nameconfiguration property to specify the name of the MySQL table from which the verifier queries will loaded.
Add configuration property
human-readable.log-fileto allow human-readable verification results to be logged into the specified file instead of
Rename configuration properties
Remove configuration properties
RowExpressionto SPI, to allow passing resolved expressions to connectors.